二變量 符號檢定法에 關한 硏究
A study of bivariate sign test
광주보건전문대학논문집 1994년 19권 1호 p.177 ~ 185
소속 상세정보
송정/Song J
KMID : 0615219940190010177
Abstract
We introduced several Bivariate Sign Tests by Hodges, Blumen and Lee in this paper.
In the bivariate normal distribution with mean vector ( μ_χ , μ_y ) and unit variance, we compared powers of some Bivariate Sign Tests to test null hypothesis H_0 against H_1 by changing the significance level α with the correlation coefficient ρ, Mahalanobis distance Δ^2, and the direction of shift θ in alternative hypothesis.
Because Hotelling´s Τ^2 test has the same power under the same Mahalanobis distance Δ^2, by changing the direction of shift in alternative hypothesis, there is a significance to compare the Τ^2 test with other Bivariate Sign Tests.
As a result, Lee´s Bivariate Sign Test is more poweful than other Bivariate Sign Tests when the direction of shift θ in alternative hypothesis is about 45?? for positive ρ. But Lee´s Bivariate Sign Test does not have a correlation with the significance level and number of samples. And Hodges Bivariate Sign Test has some strength when the significance level is high.
키워드
원문 및 링크아웃 정보
등재저널 정보